Augmented Reality
From Open Source Ecology
Basics
- Often abbreviated as AR
- A "layering" of digital data similar to that used in VR or MR, but layered directly over real life
- Thus a semi-transparent "screen" is used
- Can use inside out, or outside in tracking, although inside out is most common due to the highly mobile nature of most AR applications
- Most common tracking methods are
- The core displays are easy (phone, phone and Google Cardboard type headset, dedicated AR HMD)
- The main challenges are:
- Tracking
- Software (This is changing with OpenXR though..._
- Social Acceptance (especially of bulky, goofy looking headsets being "what works", and fashionable headsets like Google Glass (which isn't TECHNICALLY ar...), and Magic Leap , being near useless )
Main OSE Use Cases
Visually guided assemblies
- Can use a headset or some sort of handheld display
- Uses Object Recognition to see where the "parts" and "tools" are, and then calculate + show what needs to be done in the current "step"
- May be hardware intensive, but is doable, especially with [AR/MR/VR Display Tethering]]
AR CAD
- Open Source AR CAD Software
- Essentially fiddling around with your hands, but it goes to cad
- May allow for multiple people, if using a very connected work enviroment, or a LARGE pc/server capable of many virtual clients
Art Displays, Games, and other Entertainment Use
- It's been done at some film festivals, but may? have less potential as VR or MR
Enterprise
- Project North Star is VERY capable hardware wise
- Still needs software, but this is developing
- The main issue is the need for a 3D Printer, Maker Experience, And part sourcing
- ALL OF WHICH a microfactory/makerspace etc could solve
- Thus this could become a small scale enterprise (or larger if long distance online sales are included in the business plan)\
Comparison of Different Tracking Methods
Outside In
Optitrack
- The most accurate system
- VERY EXPENSIVE to set up
- cheap to add props/headsets, and many can be added (leds and ping pong balls etc are all that are needed? NEEDS MORE RESEARCH)
- Warehouse scale is about the maximum size
SteamVR
- Mid Cost to Set up
- Mid cost to add additional props/headsets
- very accurate, and probably the most accurate consumer grade outside in tracking
- Large Room Scale is the current maximum size officially, yet warehouse scale is in development
- Also max number of tracked objects?
- Also easier to source OTS parts due to the consumer nature of the system
OpenVR
PC Emulation of PSVR Tracking
- Similar method to Optitrack, with semi? open source hardware and software
- Also lesser quality, and low level of development
Inside Out
- Any usable OTS options?
- Meta AR had a system
- Project North Star has one?
- varios vr headsets, but they are closed or VERY closed source
- HTC Valve Cosmos (inside out version)
- Oculus VR (post facebook aquisition) Occulus Quest, and Occulus Rift Pro (uses cloud software? and is facebook owned so is essentially the anti-open source lol
- Tracker based can work for small stuff, but suffers occlusion issues, and drift, but is usable for many small scal applications perhaps
Use Case for Build Instructionals using Markers
- FLOSS using https://www.learnopencv.com/augmented-reality-using-aruco-markers-in-opencv-c-python/ - Example using simple markers (ArUco) markers - with Python. When you see an icon, app replaces image with another image to augment information of image. OSE Use Case: building a 3D printer, aruco marker is attached to a part, and a video tells you how to build that part. This way, just with an app and marked parts - you can build an entire thing with 'self-generated' instructions. The savings here come from not needing to identify how a part goes together by looking at documentation. This requires you to (1) find and identify part; (2) follow instructions on that part. Challenges: identifying a part from many parts can be tricky if you have to dig through a bunch of parts. Following instructions can be cumbersome. Solutions with AR: part is identified automatically (pending marker). Quick on-demand, repeating instructions can be shown automatically, without you going through pages or hitting play for a video.
- Overall SWOT: good to identify parts, but you still have to put on the labels. If labels are done automatically - such as by image recognition, not marker - then we are set. Threat: cumbersome to learn unless there is a clear instructional. Also, small parts such as small screws - it's not easy to label them. Conclusion: Image Recognition + AR is the solution. *Image Recognition*
Links
Internal Links
External Links
